Microdeletion encompassing MAPT at chromosome17q21.3 is associated with developmental delay andlearning disabilityCharles Shaw-Smith1,8, Alan M Pittman2,8, Lionel Willatt3,8, Howard Martin4, Lisa Rickman1, Susan Gribble5,Rebecca Curley5, Sally Cumming4, Carolyn Dunn3, Dimitrios Kalaitzopoulos5, Keith Porter5, Elena Prigmore5,Ana C V Krepischi-Santos6, Monica C Varela7, Celia P Koiffmann7, Andrew J Lees2, Carla Rosenberg6,Helen V Firth1, Rohan de Silva2 & Nigel P Carter5

Recently, the application of array-based comparative genomichybridization (array CGH) has improved rates of detection ofchromosomal imbalances in individuals with mental retardationand dysmorphic features1–4. Here, we describe threeindividuals with learning disability and a heterozygous deletionat chromosome 17q21.3, detected in each case by array CGH.FISH analysis demonstrated that the deletions occurred asde novo events in each individual and were between 500 kband 650 kb in size. A recently described 900-kb inversion thatsuppresses recombination between ancestral H1 and H2haplotypes5 encompasses the deletion. We show that, in eachtrio, the parent of origin of the deleted chromosome 17 carriesat least one H2 chromosome. This region of 17q21.3 showscomplex genomic architecture with well-described low-copyrepeats (LCRs)5,6. The orientation of LCRs flanking the deletedsegment in inversion heterozygotes is likely to facilitate thegeneration of this microdeletion by means of non-allelichomologous recombination.

In clinical cytogenetics, the phenotypic recog-nition of microdeletion syndromes hasusually preceded the elucidation of theunderlying causative cytogenetic imbalance.Velocardiofacial syndrome, Prader-Willi syn-drome and Williams syndrome were alldescribed clinically before the cytogeneticimbalances responsible for these syndromeswere identified7. In contrast, the introductionof FISH-based screening for subtelomericchromosomal imbalances8,9 has led to thecytogenetic characterization of new syn-

dromes, such as 22q13 deletion syndrome10, in advance of theirclinical recognition. Further refinements in detection have beenprovided by the introduction of array CGH1–4. One of the individualsdescribed in our previous report2 had a de novo deletion at chromo-some 17q21.3 involving a single clone, RP5-843B9. We have identifiedtwo further individuals with deletions of the same clone, one of whomis the subject of a recent clinical report suggesting a phenotypicresemblance to Angelman syndrome11. There are some clinical simila-rities between the three affected individuals (Table 1). All had birthweights at or below the 2nd centile. Hypotonia and/or poor feedingwere noted in the neonatal period. Motor and speech developmentwere delayed. Facial features are mildly dysmorphic in each case(Fig. 1 and Table 1).

Detailed mapping by FISH with fully sequenced tiling pathBAC and P1-derived artificial chromosome (PAC) clones indicatedthat the deletion was between 500 and 650 kb in size. Each of the threecases appeared to have a deletion for the same BAC or PAC clones,making the deletions identical in size at this level of resolution

Figure 1 Clinical photographs of affected individuals. Craniofacial dysmorphic features are presented

for each case in Table 1. We obtained informed consent to publish the photographs above.

Received 15 May; accepted 10 July; published online 13 August 2006; doi:10.1038/ng1858

1University of Cambridge Department of Medical Genetics, Addenbrooke’s Hospital, Cambridge CB2 2QQ, UK. 2Reta Lila Weston Institute of Neurological Studies,University College London, 1 Wakefield Street, London, WC1N 1PJ, UK. 3Regional Cytogenetics Laboratory and 4Regional Molecular Genetics Laboratory, Addenbrooke’sHospital, Cambridge CB2 2QQ, UK. 5The Wellcome Trust Sanger Institute, Wellcome Trust Genome Campus, Hinxton, Cambridge CB10 1SA, UK. 6Department of Geneticsand Evolutionary Biology, University of Sao Paolo, PO Box 11461, 05422-970 Sao Paolo, Brazil. 7Human Genome Research Centre, University of Sao Paolo, 05508-090Sao Paolo, Brazil. 8These authors contributed equally to this work. Correspondence should be addressed to C.S.S. (css@sanger.ac.uk).

1 03 2 VOLUME 38 [ NUMBER 9 [ SEPTEMBER 2006 NATURE GENETICS

LE T TERS©

2006

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

egen

etic

s

(Supplementary Table 1 online). The centromeric breakpoint waswithin clone RP11-707O23, which gave a diminished signal on onechromosome 17 homolog (normal flanking clone RP11-798G7;deleted flanking clone CTD-3070M1). The telomeric breakpoint wasbetween clone RP5-843B9 (no signal on one chromosome 17 homo-log) and clone RP11-259 G18 (normal signals on both chromosome17 homologs). We subsequently undertook FISH studies with clonesfrom the RP11 library that had previously been shown to be specific tothe H1 and H2 haplotypes5 (Fig. 2; details of the FISH probes used inthis analysis are given in Supplementary Table 2 online, and examplesof FISH images are given in Supplementary Fig. 1 online) The H1and H2 clones on either side of the deleted region showed signals in17q24 as well as 17q21.3, consistent with the known presence ofsegmental duplications within this region6 (Supplementary Table 1)and with the LCRs indicated in Figure 2.

The previously described 900-kb inversion5 provides an explanationfor the absence of recombination in this region and for the occurrenceof two ancestral haplotypes, which have been named H1 and H2. Webuilt on existing data5,6,12 in order to construct a detailed assembly ofthe H2 haplotype in relation to a reference H1 haplotype (Universityof California Santa Cruz (UCSC) Genome Browser, May 2004assembly), depicting SNP marker and LCR order and orientation.We used the available sequence and positioning of the available H2BAC clones relative to H1 for annotation of the architectural features(Fig. 2). The differences between the H1 and H2 assemblies are farfrom that of a simple inversion: there are also substantial differencesbetween H1 and H2 in both the number and the orientation ofsegmental duplications. For example, LCR subunit 4 lies in invertedorientation on the H1 haplotype, but it has undergone an inversion

event on the H2 haplotype such that the LCRs now lie in directorientation on this background. SNP marker position and orientationdiffer on the H1 and H2 haplotypes. The H1 centromeric markersrs2696425 and rs2668643 are duplicated on both H1 and H2, bothflanking the deletion. Some SNP markers are present at one copy onH1 (rs1528072 and rs2532418) but are duplicated on H2, and othersare present at one copy in both haplotypes (rs241041 and rs199528).Markers that fall within the confines of the inversion are in reverseorder on H2 relative to H1 (rs1396862, rs916793, H1/H2 238-bpinsertion/deletion polymorphism13 and rs1468241).

We performed detailed haplotyping with respect to H1 and H2alleles for each member of each trio (Fig. 3). In trio 1, the father is anH1/H2 heterozygote, and the mother an H2/H2 homozygote. Theparental origin of the MAPT deletion is the H1/H2 father. Ofparticular interest, the chromosome carrying the deletion in trio 1 iscomposed of both H1 and H2 alleles, suggestive of an H1/H2recombination event. To our knowledge, this is the first documentedexample of recombination between H1 and H2 haplotypes. In trio 2,the father is an H2/H2 homozygote and the mother an H1/H2heterozygote. Genotypes of all markers in the affected individual areconsistent with H2. We were unable to determine the parental originof the deletion in this trio despite typing a panel of repeat markersacross the deleted region. In trio 3, the father is an H1/H2 hetero-zygote and the mother an H1/H1 homozygote. Haplotype analysisexcluded the maternal origin of the deletion. On the deleted chromo-some, both the telomeric and centromeric markers are of type H2, asseen in trio 2.

We noted that variations in genomic architecture associated withthe H1 and H2 haplotypes at 17q21.3 present possibilities for the

Table 1 Clinical features of affected individuals with the 17q21 microdeletion

Individual 1 Individual 2 Individual 3

Current age 20 years 13 years 3 years

Gender M F F

Gestation Term 38 weeks 38 weeks

Birth weight (centile) 2.72 kg (2nd–9th) 2.35 kg (0.4th–2nd) 2.56 kg (0.4th–2nd)

Perinatal Hypotonic postnatally Hypotonic postnatally. Extended breech

presentation. Bilateral subluxation of

hips

Hypotonic postnatally

Feeding and speech Poor feeding; required nasogastric tube

feeding from 4 weeks of age

Oral dyspraxia

Slow feeding with weak suck

Oral dyspraxia

Slow feeding with poor suck

Excessive chewing/mouthing

Age at sitting 16 months 8 months 8 months

Age at walking 24 months 21 months 19 months

Learning disability Special school for children with moderate to

severe learning disability

Special school for children with

moderate learning disability

Motor and speech delay

Behavior Unremarkable Unremarkable Frequent laughing

Head circumference (percentile) 25th–50th at 17 years 9th –25th at 13 years 25th–50th at 2 years

Height (percentile) o0.4th at 17 years 9th–25th at 13 years 75th–90th at 2 years

Weight (percentile) 0.4th–2nd at 17 years 9th–25th at 13 years 50th–75th at 2 years

Dysmorphic craniofacial features Long face, deep-set eyes, narrow nose with

bulbous tip. High palate

Long face. Submucous cleft palate and

tongue tie repair

Wide mouth, short philtrum

Other findings Pectus excavatum, two cafe au lait patches,

mild contractures of elbows and knees;

reduced visual acuity in each eye with static

mild pigmentary macular epithelial changes;

no cataract. Bilateral undescended testes.

Petit mal seizures in early childhood

Pectus excavatum, small joint

hypermobility, bilateral cataract

extraction and lens implantation

at 10 years. Mild bilateral

Single clonic seizure at 13 months,

congenital dislocation of left hip,

scoliosis, strabismus, hypermetropia

hydronephrosis

Mild bilateral hearing loss

Deep sacral dimple

NATURE GENETICS VOLUME 38 [ NUMBER 9 [ SEPTEMBER 2006 1 03 3

LET TERS©

2006

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

egen

etic

s

generation of deletions by the mechanism of non-allelic homologousrecombination (NAHR), as has been described for other disorders14.The LCR subunit 4 maps closest to the deletion breakpoints on H1and H2 (Fig. 2). In H1/H2 heterozygotes, the change in orientation ofLCR 4 on the H2 background means that NAHR between thesesubunits is possible, generating a deletion approximately 500–600 kbin size (Supplementary Fig. 2 online). LCR 4 subunits are 498%identical to one another and are largely composed of interspersedrepeating elements (58%), predominantly Alu repeats. One predictionto arise from this is that NAHR between H1 and H2 chromosomesshould give rise to a chromosome consisting of proximal H1–typealleles and distal H2–type alleles, or vice versa. Such a chromosome isindeed observed in the case of trio 1, case 3, lending support to theproposed mechanism. In trio 3, centromeric and telomeric markersare of type H2, despite an apparently heterozygous parent of origin.Possible explanations for this include a mechanism other than NAHRin this case, a conversion event involving the centromeric markers ashas been described between LCRs in other genomic locations15,16, orthe inheritance of non-paternal alleles. In trio 2, it is more likely thatthe NAHR event could have occurred in the homozygous H2 father.

On the H2 background, LCRs 4 and 4a are directly oriented, so NAHRbetween these repeats is also possible (Supplementary Fig. 2). Incontrast, the LCRs are all inversely oriented on the H1 background,not satisfying the conditions required for NAHR; consistent withthis, no parent of origin in any of our trios is homozygous for theH1 haplotype.

Previous studies have reported chromosomal inversions occurringwith greater frequency in the parents of individuals with heterozygousdeletions. Mothers of individuals with Angelman syndrome with BP2/3-type deletions have a heterozygous inversion of the deleted region infour of six cases; this inversion is present in only 4 out of 44 controls(9%)17. Sotos syndrome is another example of a genomic disorder inwhich a commonly occurring deletion is flanked by LCRs18. Theserepeats lie in inverted orientation, with the exception of subunit C inthe proximal Sos-REP and subunit C¢ in the distal Sos-REP, whichlie in direct orientation and which have been shown to containa recombination hotspot for strand exchange during NAHR19.A heterozygous inversion of the interval between these LCRs hasbeen identified in all fathers of individuals carrying a deletion on thepaternally derived chromosome in a Japanese cohort19. Given that

40.9

798G7

94M7*

57A24

450G10300H14

162O14

559k6 401F3769P22

100C5

573G23207l10*

374N3

450L21

1070B7

396D2

84A7

rs19

9528

rs15

2807

2

rs25

3241

8rs

2410

41rs

2668

643

rs26

9642

5

rs13

9686

2

rs91

6793

rs14

6824

1

rs15

2807

2

rs25

3241

8

rs26

6864

3

rs26

9642

5

Rep

eat-t

rsH

1/H

2 in

s/de

l

1

1

2

2

2 8 3

3

2

2 8

83

3 8 2 3 4 4a 4a

44

4

CRHR1

CRHR1

MAPT

MAPT

339E12

707023*

329D18

413P22

∆

∆

256H10

256F16

244K17 201P9170C3

219F9

141H9

133E17*

259G18

562H3

41.0 41.1 41.2 41.3 42.1 42.2 42.341.4 41.5 41.6 41.7 41.8 41.9 42.0

rs26

9642

5

rs26

6864

3

rs24

1041

rs13

9686

2

rs91

6793

Rep

eat-t

rsH

1/H

2 in

s/de

l

rs14

6824

1

rs15

2807

2

rs25

3241

8

rs26

6864

3

rs26

9642

5

rs19

9528

Mb

Marker

H1 BAC clones

LCRs

LCRs

MAPT deletion

MAPT deletion

H1D2

H2

H2 BAC clones

Marker

Figure 2 MAPT haplotypes at 17q21.31. The H2 assembly is that of ref. 5, taking into account the LCR structure of ref. 6 and the SNP markers used

in this study for extended haplotype analysis in the triads (trio 1, trio 2 and trio 3). Shown are the different orientation of the SNPs, LCRs and two genes:

microtubule associated protein tau (MAPT) and corticotrophin releasing hormone receptor 1 (CRHR1). The H1 assembly is that of UCSC (May 2004) and

is for reference only, as H1 haplotypes show considerable structural diversity and copy number polymorphisms (CNPs) in the 3¢ LCR cluster. However, the

orientation is unlikely to change, because the CNPs are duplications and triplications. The differences between the H1 and H2 assemblies are far from that

of a simple inversion: other structural differences include segmental duplications and smaller inversion duplications (such as 4a). The orientation of LCR pair

4 on H1 is switched from opposing to tandem and is duplicated on the H2 background. SNP marker orientation differs between H1 and H2. Also shown are

results of FISH analysis using H1- and H2-specific clones from the RP11 library, as detailed in Supplementary Table 2. Clones have been assigned to H1 or

H2 haplotypes according to ref. 5. Green rectangle indicates a signal of equal intensity present on each homolog. Blue rectangle represents a signal on one

homolog of diminished intensity compared with other homolog. Red rectangle indicates a signal on one homolog only (deleted on the other homolog).

FISH images for clones with asterisks are given in Supplementary Figure 1. Repeat-t: MAPT tetranucleotide repeat allele.

1 03 4 VOLUME 38 [ NUMBER 9 [ SEPTEMBER 2006 NATURE GENETICS

LET TERS©

2006

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

egen

etic

s

microdeletions are recognized to make a greater contribution thanmutations in the NSD1 gene to the pathogenesis of Sotos syndrome inJapanese relative to Western cohorts19, it might be of interest, byanalogy with the present study, to examine the world-wide populationincidence of the genomic inversion in the Sotos syndrome region andits impact on the ratio of cases with a microdeletion versus those withan NSD1 mutation.

The deletion encompasses the gene encoding microtubule-associatedprotein tau (MAPT), and, to the best of our knowledge, these casesrepresent the first reports of haploinsufficiency of MAPT in humans.Gain-of-function mutations in MAPT cause frontotemporal dementiawith parkinsonism linked to chromosome 17 (FTDP-17)20,21, andcommon genetic variation of MAPT is associated with progressivesupranuclear palsy22 and corticobasal degeneration and Alzheimer’sdisease23. The possible role of MAPT in development has been studiedin mice lacking the tau gene (tau�/�)24. These mice are phenotypicallyand developmentally normal. The sole identified histological abnorm-ality is a reduction in microtubule density and stability in some small-caliber axons. More detailed phenotypic analysis of these mice hassubsequently been reported25. Muscle weakness in the wire-hangingtest, hyperactivity in a novel environment and impairment of con-textual fear conditioning were found, suggesting that a possible role forMAPT haploinsufficiency in learning disability in humans meritsfurther investigation. The deleted region contains other genes: cortico-trophin releasing hormone receptor 1 (CRHR1), intramembraneprotease 5 (NM_175882, also known as IMP5) and the predictedgenes NP_689679.1, NP_787078.1 and KIAA1267. Of these, there issome evidence for a role for CHRH1 in central nervous systemdevelopment26, although no data are available on the possible impactof haploinsufficiency of this gene on that process.

Microscopic chromosomal deletions involving 17q21.3 have beenreported before27, but these have been much larger than the sub-microscopic deletion reported here, and the clinical phenotypes havebeen quite different, with major congenital malformations such asesophageal atresia and congenital heart defect. The DECIPHERdatabase (http://decipher.sanger.ac.uk) was set up in order to facilitateand expedite the identification of new disorders in clinical cytoge-netics and has proved instrumental in identifying the overlappinggenotypic and phenotypic features of the three cases reported in thepresent study. Low birth weight (0.4th–9th centile), neonatal hypoto-nia, poor feeding in infancy and oromotor dyspraxia together withmental retardation seem to be common features of deletion of17q21.3. When occurring together, these features should promptconsideration of this diagnosis. Further cases are needed in order todefine the full phenotype of this disorder and its evolution with age.For genetic counseling, the presence of an inversion in the transmit-ting parent might be considered a risk factor for deletion in theoffspring. This inversion may prove to be a necessary factor fordeletion to occur, but it is unlikely to be sufficient, given the relativelyhigh frequency of the inversion in the populations in which it occurscompared with the much lower frequency of the deletion. A predictionfrom the present study is that if H1/H2 heterozygotes and possiblyH2/H2 homozygotes are predisposed to undergo deletions at 17q21.3,then the incidence of these deletions should be broadly restricted tothose populations in which the H2 haplotype occurs, namely those ofIceland, Europe and the Middle East5,28. Again, investigation of thispossibility will require analysis of additional cases. Furthermore, afollow-up study in which the parents of individuals with this newmicrodeletion syndrome are genotyped with respect to H1/H2 wouldclearly be worthwhile.

1 2

2283

2222

2

22

2222

2

2283

2222

2

22

2222

2

2283

2222

2

22

2222

2

2283

2222

2

22

2222

2

2283

2222

2

22

2222

2

2283

2222

2

22

2222

2

2299

2222

2

22

2222

2

2255

2222

2

22

2222

2

2255

2222

2

22

222

2222

1

--

---1

-

---

22

2

2

--

---2

-

---

22

2

2

--

---2

-

---

111

111111--1

325

1111

111111--1

312

1111

111111--1

308

1111

111111--1

308

1111

111111--1

287

1111

111111--1

312

∆ ∆ ∆

Trio 1 Trio 2 Trio 3

rs2696425rs2668643rs1396862

rs1468241

rs2532418rs2668643rs2696425rs2532418rs1528072rs199528

rs1528072

rs916793repeat-t

H1/H2 ins/del

rs2696425rs2668643rs1396862

rs1468241

rs2532418rs2668643rs2696425rs2532418rs1528072rs199528

rs1528072

rs916793repeat-t

H1/H2 ins/del

rs2696425rs2668643rs2532418

repeat-t

rs1396862rs2696425rs2668643rs2532418rs1528072rs199528

rs916793

rs1528072rs1468241

H1/H2 ins/del

rs2696425rs2668643rs2532418

repeat-t

rs1396862rs2696425rs2668643rs2532418rs1528072rs199528

rs916793

rs1528072rs1468241

H1/H2 ins/del

rs2696425rs2668643rs2532418

repeat-t

rs1396862rs2696425rs2668643rs2532418rs1528072rs199528

rs916793

rs1528072rs1468241

H1/H2 ins/del

rs2696425rs2668643rs1396862

repeat-t

rs2532418rs2668643rs2696425rs2532418rs1528072rs199528

rs1528072

rs916793

rs1468241H1/H2 ins/del

rs2696425rs2668643rs1396862

repeat-t

rs2532418rs2668643rs2696425rs2532418rs1528072rs199528

rs1528072

rs916793

rs1468241H1/H2 ins/del

rs2696425rs2668643rs2532418

repeat-t

rs1396862rs2696425rs2668643rs2532418rs1528072rs199528

rs916793

rs1528072rs1468241

H1/H2 ins/del

rs2696425rs2668643rs1396862

repeat-t

rs2532418

rs2696425rs2668643

rs2532418rs1528072rs199528

rs1528072

rs916793

rs1468241H1/H2 ins/del

rs2696425rs2668643rs2532418

repeat-t

rs1396862rs2696425rs2668643rs2532418rs1528072rs199528

rs916793

rs1528072rs1468241

H1/H2 ins/del

rs2696425rs2668643rs1396862

repeat-t

rs2532418rs2668643rs2696425

n/an/a

rs199528

rs1528072

rs916793

rs1468241H1/H2 ins/del

rs2696425rs2668643rs1396862

repeat-t

rs2532418rs2668643rs2696425

n/an/a

rs199528

rs1528072

rs916793

rs1488241H1/H2 ins/del

Figure 3 Extended MAPT haplotypes in the three trios. The figure is a depiction of the genotyping data in Supplementary Table 4 online and the FISH data

in Supplementary Table 1. The phased parental haplotypes (boxed) have been constructed in accordance with strict H1/H2 ruling from the SNP marker

genotypes, and the deletion haplotypes of the affected offspring have been inferred (unboxed) from the genotype data in Supplementary Table 2. Markers in

faint italics cannot reliably be used to infer haplotypes of the deleted chromosome because they lie within the confines of the deleted region. Dotted line (D)

corresponds to the deleted region established by FISH. Included in the haplotype analysis are genotypes of the size of a mixed tetranucleotide repeat

(repeat-t) situated in MAPT. Repeat-t distinguishes between the parental chromosomes in trio 1 and trio 3 and excludes a maternal origin.

NATURE GENETICS VOLUME 38 [ NUMBER 9 [ SEPTEMBER 2006 1 03 5

LET TERS©

2006

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

egen

etic

s

METHODSPatient recruitment. Individuals with learning difficulties (mental retardation)

and dysmorphic features were studied by array CGH, with detailed selection

criteria as described2. Individual 1 has been reported as part of our previous

study2. Individual 2 was subsequently recruited into the same study from the

UK. Individual 3 was first referred to the Genetic Service of the Institute of

Bioscience (University of Sao Paulo, Brazil) for evaluation. This individual has

been described in detail in a recent publication11. We obtained ethical approval

with informed consent to participate in the study as described2,11.

Whole-genome array CGH. We performed array CGH using a whole-genome

BAC array with clones spaced at an average of 1 Mb and analyzed data as

previously described2,4.

FISH. We undertook FISH studies with fully sequenced tiling path BAC and

PAC clones and subsequently with BAC clones from the RP11 library identified

as specific to the H1 and H2 haplotypes, as previously described2. The CTD

library clones were obtained from Invitrogen. The RP5 and RP11 clones were

from the BACPAC Resources Center (http://bacpac.chori.org/order.php), with

the exception of clone RP11-707O23, which was obtained from A. Cook at the

Broad Institute.

H1/H2 marker and LCR assembly. We adapted our marker and LCR assembly

from refs. 5, 6 and 12 and tailored them to the needs of this study. The H1

assembly is essentially that of the UCSC May 2004 assembly, NCBI build 35

(H1D2 in ref. 5). The H2 assembly is derived from refs. 5 and 12. We mapped

architectural features of interest to available H2 BAC clones using standard

internet-based sequence alignment programs (BLAT, BLAST). In the case of

our SNP marker assembly, the position of each SNP marker corresponds to a

PCR product generated by our primer design. Segmental duplications or LCRs

of interest flanking the microdeletion are positioned on both the H1 and H2

backgrounds in the relative orientation to one another as they would appear on

the inverted chromosomes lying side by side.

Marker genotyping. Each of our panel of SNP markers has been used in

previous population genetic studies29 and/or has been genotyped by the

International HapMap Consortium. The markers are in perfect linkage

disequilibrium (D¢ ¼ 1, r2 ¼ 1) with each other and H1/H2 (ref. 5); they

define distinctions between the H1 and H2 clades, and this innate property of

the region essentially negates problems of phasing parental chromosomes. The

238-bp MAPT H2 deletion in intron 9 was used to determine the tau H1/H2

haplotype in the triads as previously described13. In addition, we genotyped a

further panel of informative SNPs. Oligonucleotide primer pairs were designed

using Primer3 (http://frodo.wi.mit.edu/cgi-bin/primer3/primer3_www.cgi) to

amplify by PCR the MAPT haplotype SNPs (rs2696425, rs2668643, rs2532418,

rs1528072, rs1468241, rs916793, rs1396862, rs1528072, rs199528). The reac-

tions contained 25 ng of template DNA, 1 unit of Taq DNA polymerase, 10�PCR buffer, 10 mM dNTP mix, 10 pmol of each forward and reverse primer

and 5� Q solution (Qiagen). Thermocycling was performed by touchdown

PCR. Genotypes were obtained by direct sequencing and by restriction

digestion (RFLP) as previously described29. The MAPT tetranucleotide repeat

alleles were sized by capillary electrophoresis of fluorescently labeled (FAM)

PCR products an ABI 3730 Genetic Analyzer. A further panel of dinucleotide

markers were also typed by the same method as previously described12.

Sequences of primers used in this analysis are given in Supplementary

Table 3 online.

Haplotype analysis. Genotypes obtained from the H1/H2 insertion/deletion

polymorphism, the SNPs and MAPT tetranucleotide repeat alleles were used to

construct haplotypes in the triads. SNP alleles have been converted to ‘1’ or ‘2’,

corresponding to H1 and H2 alleles, respectively, for ease of interpretation.

H1/H2 parental and affected offspring ‘wild-type’ haplotypes were constructed

in accordance with H1/H2 doctrine. The deleted haplotypes of affected

offspring have been inferred where possible with respect to the size and

location of the deletion derived from FISH, assuming that no additional

structural alterations in addition to the deletion have taken place. Included

in the haplotype analysis are the allele sizes of the MAPT tetranucleotide repeat

alleles to distinguish between H1 and H2 parental chromosomes.

Note: Supplementary information is available on the Nature Genetics website.

ACKNOWLEDGMENTSThis work was funded by the Wellcome Trust. A.P., R.d.S., and A.J.L. are fundedby the Reta Lila Weston Trust for Medical Research. This work is also supportedby a grant from the UK Medical Research Council (MRC) to R.d.S. The Brazilianauthors are supported by the State of Sao Paulo Foundation for Research(FAPESP) and the Brazilian National Foundation for Research (CNPq). Arrayswere printed by the Microarray Facility of the Wellcome Trust Sanger Institute.The authors thank L. Raymond for use of laboratory facilities, J. Cox forbioinformatics advice and G. Parkin and E. Kerr for technical support. Theauthors wish to thank the affected individuals and their families for contributingto this research study.

COMPETING INTERESTS STATEMENTThe authors declare that they have no competing financial interests.

Published online at http://www.nature.com/naturegenetics

Reprints and permissions information is available online at http://npg.nature.com/

reprintsandpermissions/

1. Vissers, L.E. et al. Array-based comparative genomic hybridization for the genomewidedetection of submicroscopic chromosomal abnormalities. Am. J. Hum. Genet. 73,1261–1270 (2003).

2. Shaw-Smith, C. et al. Microarray based comparative genomic hybridisation (array-CGH)detects submicroscopic chromosomal deletions and duplications in patients withlearning disability/mental retardation and dysmorphic features. J. Med. Genet. 41,241–248 (2004).

3. Cheung, S.W. et al. Development and validation of a CGH microarray for clinicalcytogenetic diagnosis. Genet. Med. 7, 422–432 (2005).

4. Rosenberg, C. et al. Array-CGH detection of micro rearrangements in mentally retardedindividuals: clinical significance of imbalances present both in affected children andnormal parents. J. Med. Genet. 43, 180–186 (2006).

5. Stefansson, H. et al. A common inversion under selection in Europeans. Nat. Genet.37, 129–137 (2005).

6. Cruts, M. et al. Genomic architecture of human 17q21 linked to frontotemporaldementia uncovers a highly homologous family of low-copy repeats in the tau region.Hum. Mol. Genet. 14, 1753–1762 (2005).

7. Cassidy, S. & Allanson, B. Management of Genetic Syndromes (Wiley-Liss, Hoboken,New Jersey, 2005).

8. Flint, J. et al. The detection of subtelomeric chromosomal rearrangements in idiopathicmental retardation. Nat. Genet. 9, 132–140 (1995).

9. Knight, S.J. et al. Subtle chromosomal rearrangements in children with unexplainedmental retardation. Lancet 354, 1676–1681 (1999).

10. Phelan, M.C. et al. 22q13 deletion syndrome. Am. J. Med. Genet. 101, 91–99(2001).

11. Varela, M.C. et al. A 17q21.31 microdeletion encompassing the MAPTgene in a mentally impaired patient. Cytogenet. Genome Res. 114, 89–92(2006).

12. Hardy, J. et al. Evidence suggesting that Homo neanderthalensis contributedthe H2 MAPT haplotype to Homo sapiens. Biochem. Soc. Trans. 33, 582–585(2005).

13. Baker, M. et al. Association of an extended haplotype in the tau gene with progressivesupranuclear palsy. Hum. Mol. Genet. 8, 711–715 (1999).

14. Stankiewicz, P. & Lupski, J.R. Genome architecture, rearrangements and genomicdisorders. Trends Genet. 18, 74–82 (2002).

15. Hurles, M.E. Gene conversion homogenizes the CMT1A paralogous repeats. BMCGenomics 2, 11 (2001).

16. Lopes, J. et al. Homologous DNA exchanges in humans can be explained by the yeastdouble-strand break repair model: a study of 17p11.2 rearrangements associated withCMT1A and HNPP. Hum. Mol. Genet. 8, 2285–2292 (1999).

17. Gimelli, G. et al. Genomic inversions of human chromosome 15q11-q13 in mothers ofAngelman syndrome patients with class II (BP2/3) deletions. Hum. Mol. Genet. 12,849–858 (2003).

18. Kurotaki, N. et al. Fifty microdeletions among 112 cases of Sotos syndrome:low copy repeats possibly mediate the common deletion. Hum. Mutat. 22, 378–387(2003).

19. Visser, R. et al. Identification of a 3.0-kb major recombination hotspot in patients withSotos syndrome who carry a common 1.9-Mb microdeletion. Am. J. Hum. Genet. 76,52–67 (2005).

20. Hutton, M. et al. Association of missense and 5¢-splice-site mutations in tau with theinherited dementia FTDP-17. Nature 393, 702–705 (1998).

21. D’Souza, I. et al. Missense and silent tau gene mutations cause frontotemporaldementia with parkinsonism-chromosome 17 type, by affecting multiple alternativeRNA splicing regulatory elements. Proc. Natl. Acad. Sci. USA 96, 5598–5603(1999).

22. Pittman, A.M. et al. Linkage disequilibrium, fine mapping and haplotype associationanalysis of the tau gene in progressive supranuclear palsy and corticobasal degenera-tion. J. Med. Genet. 42, 837–846 (2005).

1 03 6 VOLUME 38 [ NUMBER 9 [ SEPTEMBER 2006 NATURE GENETICS

LET TERS©

2006

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

egen

etic

s

23. Myers, A.J. et al. The H1c haplotype at the MAPT locus is associated with Alzheimer’sdisease. Hum. Mol. Genet. 14, 2399–2404 (2005).

24. Harada, A. et al. Altered microtubule organization in small-calibre axons of micelacking tau protein. Nature 369, 488–491 (1994).

25. Ikegami, S., Harada, A. & Hirokawa, N. Muscle weakness, hyperactivity and impairmentin fear conditioning in tau-deficient mice. Neurosci. Lett. 279, 129–132 (2000).

26. Bishop, G.A. & King, J.S. Corticotropin releasing factor in the embryonic mousecerebellum. Exp. Neurol. 160, 489–499 (1999).

27. Khalifa, M.M., MacLeod, P.M. & Duncan, A.M. Additional case of de novo interstitialdeletion del(17)(q21.3q23) and expansion of the phenotype. Clin. Genet. 44,258–261 (1993).

28. Evans, W. et al. The tau H2 haplotype is almost exclusively Caucasian in origin.Neurosci. Lett. 369, 183–185 (2004).

29. Pittman, A.M. et al. The structure of the tau haplotype in controls andin progressive supranuclear palsy. Hum. Mol. Genet. 13, 1267–1274(2004).

NATURE GENETICS VOLUME 38 [ NUMBER 9 [ SEPTEMBER 2006 1 03 7

LET TERS©

2006

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

egen

etic

s